The complex network of interconnected cellular signals produced in response to changes in the human body provides a vast amount of interesting and valuable information that could inform the development of more effective medical treatments for many diseases. In peripheral immune cells, these signals can be observed and quantified using a number of tools, including cell profiling techniques.
Single-cell profiling techniques, such as polychromatic flow and mass cytometry, have improved significantly in recent years and could now be used theoretically to obtain detailed immune profiles of patients with a range of symptoms.
However, the limited sample sizes of previous studies and the high dimensionality of patient data collected so far increase the chances of false-positive findings, which in turn lead to uncertain immune profiles.
Conducting studies in larger groups of patients could improve the effectiveness of these cell profiling techniques, allowing medical researchers to better understand the patterns associated with certain diseases. However, collecting data from many patients can be both costly and time consuming.
How the algorithm works in the study of some diseases
Thus, researchers at Stanford University School of Medicine have recently developed the immunological Elastic-Net (iEN), a machine learning model that predicts cellular responses based on mechanistic immunological knowledge.
One paper published in Nature Machine Intelligence, they demonstrated that incorporating this immunological knowledge into the prediction processes of their model increased the predictive power of both small and large patient data sets.
“Our methodology allows us to use previous studies to increase the accuracy of our models without enrolling additional patients,” Nima Aghaeepour, one of the researchers who led the study with Anthony Culos, Martin Angst and Brice Gaudilliere, told TechXplore. “A key advantage of our method is that it does not limit the nature of data-driven models. In cases where the data collected do not agree with previous knowledge, our algorithm is allowed to reduce the importance of previous knowledge and instead focus on raw data if it proves to be the stronger solution.
In scenarios where medical researchers have to consider a large number of dimensions, various features can be just as valuable for making predictions. Therefore, instead of throwing out variables that are inconsistent with previous immunological data, the machine learning algorithm developed by Aghaeepour and colleagues selects all immune characteristics that they consider to have strong predictive value and relevance.
To date, researchers have evaluated the performance of their machine learning algorithm in three independent studies. In all these studies, they found that their model could predict clinically relevant results based on both simulated data and mass cytometry data generated from patients’ blood.
How recovery after Alzheimer’s and Parkinson’s would help
“In our paper, we include two real-world clinical examples in which the iEN line has increased our accuracy in modeling pregnancy and periodontal disease,” said Aghaeepour. “We have several other interesting use cases that we can’t wait to see published, including postoperative recovery, Alzheimer’s disease and Parkinson’s disease.”
In the future, the machine learning platform developed by Aghaeepour and his colleagues could help to study the many diseases, medical conditions and neurological disorders. Data used by researchers and the iEN algorithm are available online, so that it could soon be accessed and used by other research teams around the world.
“We are now working on developing versions of the algorithm that are applicable to other types of biological data sets,” said Aghaeepour. “A prime example in this regard are the studies in which several technologies are used simultaneously to profile the immune system. We believe that these data sets provide unique opportunities for coding prior knowledge into machine learning algorithms. ”